OverviewRisksControlsMapReference

Search the Atlas

Search risks, controls, and glossary terms

Agentic AI risk reference

AI Risk Atlas

A governance command center for 77 canonical agentic risks and 157 controls — mapped by severity, likelihood, control depth, and eight risk frameworks including the IBM AI Risk Atlas.

Agentic · 77Non‑agentic GenAI · 64
Explore risks Browse controls
0
Canonical risks
11 governance domains
0
Critical-rated
45 high severity
0
Agentic-specific
new or amplified by agents
0
Mapped controls
avg 4.9 per risk

Inherent risk matrix

Severity × likelihood. Select a cell to drill into those risks.

Low
Medium
High
Critical
10score 413score 87score 12
High
4score 313score 622score 9
Medium
2score 22score 44score 6
Likelihood →
CriticalHighMediumLow

Priority register

Highest inherent-risk items.

View all
  1. 112Objective Substitution & Proxy GamingAI System Safety4 ctrl
  2. 212A2A Trust ExploitationMulti-Agent Security4 ctrl
  3. 312Persistent Memory CorruptionMemory & Knowledge Governance5 ctrl
  4. 412External Goal ManipulationSecurity & Identity6 ctrl
  5. 512Indirect Prompt InjectionSecurity & Identity6 ctrl
  6. 612MCP Tool Description Poisoning (TPA-D)Tool & Affordance Risk7 ctrl
  7. 712MCP Resource & Prompt Primitive InjectionSecurity & Identity5 ctrl
  8. 89Unintended Environmental ImpactAI System Safety3 ctrl

Domain coverage map

Risk load and average control depth per governance domain.

Accountability & Governance
18
10 agentic · 8 foundational5.7 ctrl/risk
Agent Autonomy
9
9 agentic6.4 ctrl/risk
AI System Safety
8
1 crit · 8 agentic3.5 ctrl/risk
Content Safety & Integrity
8
1 crit · 8 foundational4.4 ctrl/risk
Data & Input Integrity
7
7 foundational5.1 ctrl/risk
Emergent Behavior
1
1 agentic2.0 ctrl/risk
Fairness & Bias
6
6 foundational6.3 ctrl/risk
Memory & Knowledge Governance
4
1 crit · 4 agentic4.8 ctrl/risk
Model Accuracy & Reliability
8
2 crit · 8 foundational9.3 ctrl/risk
Multi-Agent Security
11
1 crit · 11 agentic3.3 ctrl/risk
Physical & Scientific Risk
3
3 agentic3.7 ctrl/risk
Privacy & Data Risk
12
1 crit · 5 agentic · 7 foundational5.0 ctrl/risk
Security & Identity
18
6 crit · 9 agentic · 9 foundational5.1 ctrl/risk
Societal & Economic Risk
11
6 agentic · 5 foundational4.2 ctrl/risk
Tool & Affordance Risk
11
1 crit · 11 agentic6.5 ctrl/risk
Transparency & Explainability
6
6 foundational5.3 ctrl/risk

Capability‑pattern applicability

The 77 risks span agentic deployment patterns and foundational (inherited) risks — every risk and control is tagged by pattern so you can scope to your architecture.

68
Agentic‑specific
new or amplified by agents
9
Foundational
traditional / generative, inherited
By pattern scope
Pattern-agnostic (machinery)57
Pattern-specific13
Pattern-agnostic (content)5
Agentic deployment1
Multi-Agent1
By agentic deployment pattern
Autonomous / Long-Horizon26
Single-Agent Tool Use16
Multi-Agent13
Foundational (inherited)11
MCP / Tool-Ecosystem7
Memory-Augmented4

Framework cross-reference coverage

How many of the 77 risks carry a mapping to each major taxonomy.

EU AI Act77/77 · 100%
DORA40/77 · 52%
MIT AI Risk Repository77/77 · 100%
IBM AI Risk Atlas77/77 · 100%
NIST ARIA77/77 · 100%
OWASP77/77 · 100%
MITRE ATLAS75/77 · 97%
MAESTRO77/77 · 100%
Cisco AI Risk72/77 · 94%
Palo Alto AI Risk56/77 · 73%
Risk explorer

Filter 77 risks by domain, severity, and likelihood.

Open
Control catalogue

157 controls with owners, thresholds, and cadence.

Open
Reference

Glossary, framework sources, and overlap register.

Open

Compiled from leading AI risk frameworks — including the IBM AI Risk Atlas, MIT AI Risk Repository, NIST AI 600-1, ISO 42001, OWASP LLM/Agentic, MITRE ATLAS, Cisco AI Defense, and Palo Alto Prisma AIRS. Inherent-risk scores are severity × likelihood; control depth reflects mapped agentic and non-agentic controls. Expert‑elicited for governance reference only.